The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and process external data from conversation history files, which could contain malicious prompts or instructions from past untrusted sessions.\n- Ingestion points: Local history logs and session files as defined in the 'Data Locations' section (e.g., ~/.codex/history.jsonl, session JSONL files).\n- Boundary markers: Absent; the skill does not provide delimiters or instructions for the agent to ignore potentially adversarial content within the history data.\n- Capability inventory: The skill requires access to the local file system to search, read, and extract content from specific tool-related directories and SQLite databases.\n- Sanitization: Absent; there are no steps defined to validate or sanitize the historical content before it is summarized or used to resume work.

codex-history